A high-frequency data processing method based on SAS

Abstract

The application discloses a high-frequency data processing method based on SAS, high-frequency data is collected through an SAS system, the collected high-frequency data is preprocessed, then a statistical analysis function of the SAS system is used to extract data features in the high-frequency data after preprocessing, the data features corresponding to the high-frequency data are displayed to staff, data analysis labels input by the staff are acquired, data analysis labels corresponding to each high-frequency data are obtained, a deep learning algorithm is used to learn data relations according to the data features corresponding to the high-frequency data and the data analysis labels corresponding to the high-frequency data, a high-frequency data processing model is acquired, and finally, the high-frequency data can be processed according to the high-frequency data processing model, so that the processing efficiency and accuracy of the high-frequency data are effectively improved, and the technical problem caused by manual processing in the prior art is solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of data processing, and in particular relates to a high-frequency data processing method based on SAS. Background Art

[0002] With the rapid development of financial markets, high-frequency data has become a crucial foundation for financial analysis and decision-making. However, the processing and analysis of high-frequency data faces challenges such as large data volumes, high real-time requirements, and complex data features. Traditional data processing methods, which primarily rely on manual analysis, are unable to meet the requirements for efficient and accurate analysis. Therefore, how to effectively process and analyze high-frequency data has become a pressing issue. Summary of the Invention

[0003] The present invention provides a high-frequency data processing method based on SAS, which is used to solve the problems of low efficiency and inaccurate analysis caused by manual data processing in the prior art.

[0004] A high-frequency data processing method based on SAS, comprising:

[0005] Collect high-frequency data through the SAS system and pre-process the collected high-frequency data to obtain the pre-processed high-frequency data; the data type of the high-frequency data is specified by the staff;

[0006] The statistical analysis function of the SAS system is used to extract the data features of the high-frequency data after preprocessing and obtain the data features corresponding to the high-frequency data;

[0007] Display the data features corresponding to the high-frequency data to the staff, obtain the data analysis labels input by the staff, and obtain the data analysis labels corresponding to each high-frequency data;

[0008] Based on the data features and data analysis labels corresponding to high-frequency data, a deep learning algorithm is used to learn data relationships and obtain a high-frequency data processing model.

[0009] The high-frequency data processing model is deployed in the cloud computing center. After collecting the data features of real-time high-frequency data through the SAS system, it is transmitted to the cloud computing center for processing, and the data analysis tags corresponding to the data features of the real-time high-frequency data are identified to obtain the high-frequency data processing results.

[0010] Furthermore, it also includes:

[0011] When the high-frequency data processing result meets the preset alarm condition, data warning information is generated and transmitted to the equipment designated by the staff.

[0012] Furthermore, high-frequency data is collected through the SAS system, and the collected high-frequency data is preprocessed to obtain the preprocessed high-frequency data, including:

[0013] Based on the data types specified by the staff, high-frequency data is collected through the SAS system;

[0014] The collected high-frequency data is cleaned, denoised and standardized to obtain the pre-processed high-frequency data.

[0015] Furthermore, based on the data features and data analysis labels corresponding to the high-frequency data, a deep learning algorithm is used to learn data relationships and obtain a high-frequency data processing model, including:

[0016] Construct a convolutional neural network or a recurrent neural network to obtain the neural network to be trained;

[0017] Initializing a population based on the hyperparameters of the neural network to be trained;

[0018] For each individual in the population, the data features corresponding to the high-frequency data are used as input, and the data analysis labels corresponding to the high-frequency data are used as the expected output to obtain the fitness of the individual.

[0019] Determine the individual with the largest fitness as the best individual and the individual with the smallest fitness as the worst individual;

[0020] According to the optimal individual, an adaptive learning strategy is used to select the initial position of each individual and determine the individual after the initial position selection;

[0021] According to the worst individual, a neighborhood scattering search strategy is used to perform neighborhood search on each individual after the initial position selection to obtain the individual after the neighborhood search;

[0022] Adopting the position influence search strategy to conduct adaptive information interaction on the individuals after each neighborhood search, the individuals after adaptive information interaction are obtained;

[0023] A positional variation search strategy is used to perform a global variation search on each individual after adaptive information interaction to obtain the individual after the global variation search.

[0024] When the individual after the global mutation search or the current number of training times meets the training end condition, the optimal individual is re-determined, and the parameters in the optimal individual are used as the final parameters of the neural network to be trained to obtain a high-frequency data processing model.

[0025] Furthermore, the data features corresponding to the high-frequency data are used as input, and the data analysis labels corresponding to the high-frequency data are used as the expected output to obtain the fitness of the individual, including:

[0026] After applying the parameters contained in the individual to the neural network to be trained, the data features corresponding to the high-frequency data are used as input to obtain the actual output of the neural network to be trained;

[0027] The data analysis label corresponding to the high-frequency data is used as the expected output, and the loss function value corresponding to the individual is obtained based on the expected output and the actual output;

[0028] Take the negative value of the loss function corresponding to the individual to obtain the fitness of the individual.

[0029] Furthermore, based on the optimal individual, an adaptive learning strategy is used to select the initial position of each individual, and the individual after the initial position selection is determined, including:

[0030]

[0031] in, represents the i-th individual in the t-th training process, i=1,2,…,N, N represents the total number of individuals in the population, Represents the individual after the initial position selection represents the optimal individual, c1 represents the first learning factor, c2 represents the second learning factor, r1 represents the first random number between (0,1), and r2 represents the second random number between (0,1). represents the kth individual in the tth training process.

[0032] Furthermore, based on the worst individual, a neighborhood scattering search strategy is used to perform neighborhood search on each individual after the initial position selection, and the individuals after the neighborhood search are obtained, including:

[0033]

[0034] in, represents the individual after the mth initial position selection during the tth training process, represents the individual after the mth initial position selection during the t-1th training process, Represents the individual after neighborhood search β1 represents the reverse search coefficient between (0,1), β2 represents the position memory coefficient between (0,0.2), and r3 represents the third random number between (0,1). Indicates the worst individual.

[0035] Furthermore, the position influence search strategy is used to perform adaptive information interaction on the individuals after each neighborhood search, and the individuals after adaptive information interaction are obtained, including:

[0036]

[0037] in, Represents the individual after the nth neighborhood search during the tth training process The d-th dimension parameter, d=1,2,…,D, D represents the total dimension of the parameter, represents the d-th dimension parameter of the individual after the n-th adaptive information interaction, r4 represents the fourth random number between (0,1), and r5 represents the fifth random number between (0,1). Represents the individual after the nth neighborhood search Randomly matched other individuals The d-th dimension parameter, T represents the preset maximum number of training times, γ nc Represents an individual With other individuals The position between the parameters, dist nc Represents an individual With other individuals The Euclidean distance between nmax Represents an individual The maximum Euclidean distance between each individual and all other individuals, dist nmin Represents an individual The minimum Euclidean distance to all other individuals.

[0038] Furthermore, a positional variation search strategy is used to perform a global variation search on each individual after adaptive information interaction, and the individuals after the global variation search are obtained, including:

[0039]

[0040] in, represents the individual after the s-th adaptive information interaction during the t-th training process, Represents the individual after global mutation search Levy represents the random Levy flight factor, α represents the position variation search control factor, Indicates that except for individual Random individuals other than Represents a newly generated random individual, e represents a natural constant, π represents the ratio of pi, r6 represents the sixth random number between (0,1), sin represents the sine function, and T represents the preset maximum number of training times.

[0041] Furthermore, the high-frequency data processing model is deployed in the cloud computing center. After the data features of the real-time high-frequency data are collected through the SAS system, they are transmitted to the cloud computing center for processing. The data analysis tags corresponding to the data features of the real-time high-frequency data are identified to obtain the high-frequency data processing results, including:

[0042] Deploy high-frequency data processing models in cloud computing centers;

[0043] Collect real-time high-frequency data through the SAS system, and extract data features of the real-time high-frequency data through the SAS system;

[0044] The data features of real-time high-frequency data are transmitted to the cloud computing center, and the cloud computing center schedules the deployed high-frequency data processing model to identify the received data features and obtain the output of the high-frequency data processing model;

[0045] A high-frequency data processing result is determined according to the output of the high-frequency data processing model.

[0046] The present invention provides a SAS-based high-frequency data processing method, which collects high-frequency data through a SAS system and pre-processes the collected high-frequency data. Then, the statistical analysis function of the SAS system is used to extract data features from the pre-processed high-frequency data. The data features corresponding to the high-frequency data are displayed to staff, and data analysis labels input by the staff are obtained to obtain data analysis labels corresponding to each high-frequency data. According to the data features corresponding to the high-frequency data and the data analysis labels corresponding to the high-frequency data, a deep learning algorithm is used to learn data relationships to obtain a high-frequency data processing model. Finally, the high-frequency data can be processed according to the high-frequency data processing model, thereby effectively improving the processing efficiency and accuracy of the high-frequency data and solving the technical problems caused by manual processing in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0048] Figure 1 The present invention provides a flowchart of a high-frequency data processing method based on SAS.

[0049] The above drawings illustrate specific embodiments of the present invention, which will be described in more detail below. These drawings and the accompanying description are not intended to limit the scope of the present invention in any way, but rather to illustrate the concept of the present invention to those skilled in the art by reference to specific embodiments. DETAILED DESCRIPTION

[0050] Exemplary embodiments will be described in detail herein, examples of which are illustrated in the accompanying drawings. In the following description, when referring to the drawings, like numbers in different figures represent like or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments are not intended to represent all possible embodiments consistent with the present invention. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present invention, as detailed in the appended claims.

[0051] The embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0052] As Figure 1 shown, the embodiment of the present application provides a high-frequency data processing method based on SAS, comprising:

[0053] S1, collecting high-frequency data through a SAS (Statistics Analysis System) system, and preprocessing the collected high-frequency data to obtain preprocessed high-frequency data; wherein the data type of the high-frequency data is specified by a staff member;

[0054] The purpose of preprocessing the collected high-frequency data is to make the data more regular and easy to be identified, so various existing technologies can be used to preprocess the collected high-frequency data to improve the identifiability and regularity of the data. The data type of the high-frequency data is specified by the staff member, which can assist the staff member to realize the self-determined high-frequency data processing task, thereby improving the processing efficiency and accuracy of the high-frequency data.

[0055] S2, using the statistical analysis function of the SAS system to extract the data features in the preprocessed high-frequency data to obtain the data features corresponding to the high-frequency data;

[0056] The statistical analysis function of the SAS system can extract the data features in the high-frequency data, making the data more easily identified, thereby improving the data recognition effect.

[0057] S3, displaying the data features corresponding to the high-frequency data to the staff member, and obtaining the data analysis label input by the staff member to obtain the data analysis label corresponding to each high-frequency data;

[0058] The data type of the high-frequency data is specified by the staff member (for example, a plurality of market high-frequency data can be determined by the staff member, or high-frequency running data corresponding to a certain device), after determining the data features, the high-frequency data and the corresponding data features can be displayed to the staff member, the staff member only needs to process the data once at the beginning (i.e. high-frequency data), and then the automatic processing can be realized, thereby improving the processing efficiency of the staff member on the high-frequency data and reducing the misprocessing caused by manual processing.

[0059] S4, using a deep learning algorithm to learn the data relationship according to the data features corresponding to the high-frequency data and the data analysis label corresponding to the high-frequency data to obtain a high-frequency data processing model;

[0060] Deep learning algorithms have the ability to learn data relationships. After learning data relationships through deep learning algorithms, a high-frequency data processing model with data processing capabilities can be obtained. This high-frequency data processing model can complete the data recognition tasks specified by the staff (that is, the recognition tasks between high-frequency data and data analysis labels).

[0061] S5. Deploy the high-frequency data processing model in the cloud computing center, collect the data features of the real-time high-frequency data through the SAS system, and transmit it to the cloud computing center for processing, identify the data analysis tags corresponding to the data features of the real-time high-frequency data, and obtain the high-frequency data processing results.

[0062] By deploying the high-frequency data processing model in the cloud computing center, we can leverage the powerful computing power of the cloud computing center to quickly identify data, while collaborating with the SAS system to assist workers in improving data recognition efficiency.

[0063] In an embodiment of the present invention, the following further comprises:

[0064] When the high-frequency data processing result meets the preset alarm condition, data warning information is generated and transmitted to the equipment designated by the staff.

[0065] The result of high-frequency data processing refers to a data classification label. Therefore, the preset alarm condition can be set to a specific data classification label (set by the staff according to the actual situation). When the data classification label corresponding to the identified real-time high-frequency data is the same as any label in the preset alarm condition, a data warning information can be generated.

[0066] In an embodiment of the present invention, high-frequency data is collected through a SAS system, and the collected high-frequency data is preprocessed to obtain the preprocessed high-frequency data, including:

[0067] Based on the data types specified by the staff, high-frequency data is collected through the SAS system;

[0068] The collected high-frequency data is cleaned (such as removing missing values), denoised (such as removing outliers), and standardized (such as numerical processing and normalization processing) to obtain high-frequency data after preprocessing.

[0069] In an embodiment of the present invention, a deep learning algorithm is used to learn data relationships based on the data features and data analysis labels corresponding to the high-frequency data to obtain a high-frequency data processing model, including:

[0070] Construct a convolutional neural network or a recurrent neural network to obtain a neural network to be trained. It is worth noting that the above neural network is only a preferred example of an embodiment of the present invention, and other neural networks can also be used as the neural network to be trained. When selecting different neural networks, the data features need to be processed into corresponding input forms.

[0071] Based on the hyperparameters of the neural network to be trained, the population is initialized; for example, the data can be initialized by a random initialization method, so that a population can be obtained, and each individual in the population includes all the hyperparameters to be trained of the neural network to be trained, and the hyperparameters to be trained can be all or part of the hyperparameters of the neural network to be trained.

[0072] For each individual in the population, the data features corresponding to the high-frequency data are used as input, and the data analysis labels corresponding to the high-frequency data are used as the expected output to obtain the fitness of the individual.

[0073] Determine the individual with the largest fitness as the best individual and the individual with the smallest fitness as the worst individual;

[0074] According to the optimal individual, an adaptive learning strategy is used to select the initial position of each individual and determine the individual after the initial position selection;

[0075] According to the worst individual, a neighborhood scattering search strategy is used to perform neighborhood search on each individual after the initial position selection to obtain the individual after the neighborhood search;

[0076] Adopting the position influence search strategy to conduct adaptive information interaction on the individuals after each neighborhood search, the individuals after adaptive information interaction are obtained;

[0077] A positional variation search strategy is used to perform a global variation search on each individual after adaptive information interaction to obtain the individual after the global variation search.

[0078] When the individual after the global variation search or the current number of training times meets the training end condition (for example, the optimal individual is re-determined based on the individual after the global variation search, and when the fitness of the optimal individual converges, the training end condition is met; or when the current number of training times is greater than or equal to the preset maximum number of training times, the training end condition is met), the optimal individual is re-determined, and the parameters in the optimal individual are used as the final parameters of the neural network to be trained to obtain a high-frequency data processing model.

[0079] Optionally, after each search, individuals may be processed as out of bounds to ensure data validity.

[0080] The prior art has the problem of poor data relationship learning effect when learning data relationship, which leads to the inability to accurately complete the data processing task specified by the staff, and therefore the embodiment of the present application provides a deep learning algorithm to improve the data processing accuracy and ultimately improve the accuracy of completing the data processing task specified by the staff.

[0081] In the embodiment of the present application, the data features corresponding to the high-frequency data are taken as the input, and the data analysis labels corresponding to the high-frequency data are taken as the expected output to obtain the fitness of the individual, including:

[0082] After the parameters contained in the individual are applied to the neural network to be trained, the data features corresponding to the high-frequency data are taken as the input to obtain the actual output in the neural network to be trained;

[0083] The data analysis labels corresponding to the high-frequency data are taken as the expected output, and the loss function value corresponding to the individual is obtained according to the expected output and the actual output;

[0084] The loss function value corresponding to the individual is taken as the negative number to obtain the fitness of the individual.

[0085] In the embodiment of the present application, according to the optimal individual, an adaptive learning strategy is adopted to select the initial position of each individual, and the individual after the initial position selection includes:

[0086]

[0087] Wherein, The i-th individual in the t-th training process is represented by i=1, 2, …, N, and N represents the total number of individuals in the population, The individual after the initial position selection is represented by The optimal individual is represented by c1, which represents the first learning factor, c2 represents the second learning factor, r1 represents the first random number between (0, 1), and r2 represents the second random number between (0, 1), The k-th individual in the t-th training process is represented by

[0088] The adaptive learning strategy provided by the embodiment of the present application can make the individual learn the information of the optimal individual, can make the algorithm always optimize, can guarantee the speed of the algorithm, and can make the individual far away from the population have a larger search step, which is not only beneficial to improving the speed of the algorithm, but also helps to improve the ability to jump out of the local optimum.

[0089] In the embodiment of the present application, according to the worst individual, a neighborhood emission search strategy is adopted to perform neighborhood search on each individual after the initial position selection to obtain the individual after the neighborhood search, including:

[0090]

[0091] in, represents the individual after the mth initial position selection during the tth training process, represents the individual after the mth initial position selection during the t-1th training process, Represents the individual after neighborhood search β1 represents the reverse search coefficient between (0,1), β2 represents the position memory coefficient between (0,0.2), and r3 represents the third random number between (0,1). Indicates the worst individual.

[0092] The neighborhood distribution search strategy provided by the embodiment of the present invention can enable individuals to memorize historical positions for neighborhood search, and at the same time use the worst individual as a reference for reverse search, thereby improving the algorithm's search capability for local areas.

[0093] In an embodiment of the present invention, a position influence search strategy is used to perform adaptive information interaction on each individual after neighborhood search to obtain an individual after adaptive information interaction, including:

[0094]

[0095] in, Represents the individual after the nth neighborhood search during the tth training process The d-th dimension parameter, d=1,2,…,D, D represents the total dimension of the parameter, represents the d-th dimension parameter of the individual after the n-th adaptive information interaction, r4 represents the fourth random number between (0,1), and r5 represents the fifth random number between (0,1). Represents the individual after the nth neighborhood search Randomly matched other individuals The d-th dimension parameter, T represents the preset maximum number of training times, γ nc Represents an individual With other individuals The position between the parameters, dist nc Represents an individual With other individuals The Euclidean distance between nmax Represents an individual The maximum Euclidean distance between each individual and all other individuals, dist nmin Represents an individual The minimum Euclidean distance to all other individuals.

[0096] The position-affected search strategy provided in the embodiment of the present invention can adjust the search step size according to the distance between two individuals. For the early and middle stages of the algorithm, it can not only improve the search speed of the algorithm, but also improve the ability to escape the local optimum to a certain extent. In the later stages of the algorithm, the search accuracy gradually increases, ensuring the convergence of the algorithm.

[0097] In the embodiment of the present invention, a position variation search strategy is used to perform a global variation search on each individual after adaptive information interaction, and the individuals after the global variation search are obtained, including:

[0098]

[0099] in, represents the individual after the s-th adaptive information interaction during the t-th training process, Represents the individual after global mutation search Levy represents the random Levy flight factor, α represents the position variation search control factor, Indicates that except for individual Random individuals other than Represents a newly generated random individual, e represents a natural constant, π represents the ratio of pi, r6 represents the sixth random number between (0,1), sin represents the sine function, and T represents the preset maximum number of training times.

[0100] The position variation search strategy provided by the embodiment of the present invention can effectively provide the global search capability of the algorithm, and is provided with a position variation search control factor, which provides a strong global search capability in the early and middle stages of the algorithm and quickly reduces the global search capability in the later stages of the algorithm to ensure the convergence of the algorithm.

[0101] The training algorithm provided by the embodiment of the present invention can effectively improve the algorithm's ability to learn data, thereby completing the data processing tasks specified by the staff more accurately and efficiently.

[0102] In an embodiment of the present invention, a high-frequency data processing model is deployed in a cloud computing center. After the data features of real-time high-frequency data are collected through a SAS system, they are transmitted to the cloud computing center for processing. The data analysis tags corresponding to the data features of the real-time high-frequency data are identified to obtain high-frequency data processing results, including:

[0103] Deploy high-frequency data processing models in cloud computing centers;

[0104] Collect real-time high-frequency data through the SAS system, and extract data features of the real-time high-frequency data through the SAS system;

[0105] The data features of real-time high-frequency data are transmitted to the cloud computing center, and the cloud computing center schedules the deployed high-frequency data processing model to identify the received data features and obtain the output of the high-frequency data processing model;

[0106] The high-frequency data processing result is determined according to the output of the high-frequency data processing model, that is, the category with the highest probability is the high-frequency data processing result.

[0107] The present invention provides a SAS-based high-frequency data processing method, which collects high-frequency data through a SAS system and pre-processes the collected high-frequency data. Then, the statistical analysis function of the SAS system is used to extract data features from the pre-processed high-frequency data. The data features corresponding to the high-frequency data are displayed to staff, and data analysis labels input by the staff are obtained to obtain data analysis labels corresponding to each high-frequency data. According to the data features corresponding to the high-frequency data and the data analysis labels corresponding to the high-frequency data, a deep learning algorithm is used to learn data relationships to obtain a high-frequency data processing model. Finally, the high-frequency data can be processed according to the high-frequency data processing model, thereby effectively improving the processing efficiency and accuracy of the high-frequency data and solving the technical problems caused by manual processing in the prior art.

[0108] Those skilled in the art will readily appreciate other embodiments of the present invention after considering the specification and practicing the invention disclosed herein. The present invention is intended to cover any variations, uses, or adaptations of the present invention that follow the general principles of the present invention and include common knowledge or customary techniques in the art not disclosed herein. It should be understood that the present invention is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and variations can be made without departing from the scope thereof. The scope of the present invention is limited only by the appended claims.

Claims

1. A high-frequency data processing method based on SAS, characterized in that: include: High-frequency data is collected through the statistical analysis system SAS, and pre-processed to obtain the pre-processed high-frequency data. The data type of high-frequency data is specified by the staff, and the high-frequency data can be high-frequency data from various markets or high-frequency operating data corresponding to a certain equipment. The statistical analysis function of the statistical analysis system SAS is used to extract the data features of the high-frequency data after preprocessing and obtain the data features corresponding to the high-frequency data; Display the data features corresponding to the high-frequency data to the staff, obtain the data analysis labels input by the staff, and obtain the data analysis labels corresponding to each high-frequency data; Based on the data features and data analysis labels corresponding to high-frequency data, a deep learning algorithm is used to learn data relationships and obtain a high-frequency data processing model. The high-frequency data processing model is deployed in the cloud computing center. After collecting the data features of real-time high-frequency data through the statistical analysis system SAS, it is transmitted to the cloud computing center for processing, and the data analysis tags corresponding to the data features of the real-time high-frequency data are identified to obtain the high-frequency data processing results. Also includes: When the high-frequency data processing result meets the preset alarm condition, data warning information is generated and transmitted to the equipment designated by the staff; Based on the data features and data analysis labels corresponding to high-frequency data, a deep learning algorithm is used to learn data relationships and obtain a high-frequency data processing model, including: Construct a convolutional neural network or a recurrent neural network to obtain the neural network to be trained; Initializing a population based on the hyperparameters of the neural network to be trained; For each individual in the population, the data features corresponding to the high-frequency data are used as input, and the data analysis labels corresponding to the high-frequency data are used as the expected output to obtain the fitness of the individual. Determine the individual with the largest fitness as the best individual and the individual with the smallest fitness as the worst individual; According to the optimal individual, an adaptive learning strategy is used to select the initial position of each individual and determine the individual after the initial position selection; According to the worst individual, a neighborhood scattering search strategy is used to perform neighborhood search on each individual after the initial position selection to obtain the individual after the neighborhood search; Adopting the position influence search strategy to conduct adaptive information interaction on the individuals after each neighborhood search, the individuals after adaptive information interaction are obtained; A positional variation search strategy is used to perform a global variation search on each individual after adaptive information interaction to obtain the individual after the global variation search. When the individual after the global mutation search or the current number of training times meets the training end condition, the optimal individual is re-determined, and the parameters of the optimal individual are used as the final parameters of the neural network to be trained to obtain a high-frequency data processing model; According to the worst individual, a neighborhood scattering search strategy is used to perform neighborhood search on each individual after the initial position selection, and the individuals after the neighborhood search are obtained, including: in, represents the individual after the mth initial position selection during the tth training process, represents the individual after the mth initial position selection during the t-1th training process, Represents the individual after neighborhood search β1 represents the reverse search coefficient between (0,1), β2 represents the position memory coefficient between (0,0.2), and r3 represents the third random number between (0,1). Indicates the worst individual.

2. The high-frequency data processing method based on SAS according to claim 1, characterized in that: High-frequency data is collected through the statistical analysis system SAS, and the collected high-frequency data is preprocessed to obtain the preprocessed high-frequency data, including: Based on the data types specified by the staff, high-frequency data are collected through the statistical analysis system SAS; The collected high-frequency data is cleaned, denoised and standardized to obtain the pre-processed high-frequency data.

3. The high-frequency data processing method based on SAS according to claim 1, characterized in that: The data features corresponding to high-frequency data are used as input, and the data analysis labels corresponding to high-frequency data are used as the expected output to obtain the fitness of the individual, including: After applying the parameters contained in the individual to the neural network to be trained, the data features corresponding to the high-frequency data are used as input to obtain the actual output of the neural network to be trained; The data analysis label corresponding to the high-frequency data is used as the expected output, and the loss function value corresponding to the individual is obtained based on the expected output and the actual output; Take the negative value of the loss function corresponding to the individual to obtain the fitness of the individual.

4. The high-frequency data processing method based on SAS according to claim 3, characterized in that: Based on the optimal individual, an adaptive learning strategy is used to select the initial position of each individual, and the individual after the initial position selection is determined, including: in, represents the i-th individual in the t-th training process, i=1,2,…,N, N represents the total number of individuals in the population, Represents the individual after the initial position selection represents the optimal individual, c1 represents the first learning factor, c2 represents the second learning factor, r1 represents the first random number between (0,1), and r2 represents the second random number between (0,1). represents the kth individual in the tth training process.

5. The high-frequency data processing method based on SAS according to claim 4, characterized in that: The position influence search strategy is used to perform adaptive information interaction on the individuals after each neighborhood search, and the individuals after adaptive information interaction are obtained, including: in, Represents the individual after the nth neighborhood search during the tth training process The d-th dimension parameter, d=1,2,…,D, D represents the total dimension of the parameter, represents the d-th dimension parameter of the individual after the n-th adaptive information interaction, r4 represents the fourth random number between (0,1), and r5 represents the fifth random number between (0,1). Represents the individual after the nth neighborhood search Randomly matched other individuals The d-th dimension parameter, T represents the preset maximum number of training times, γ nc Represents an individual With other individuals The position between the parameters, dist nc Represents an individual With other individuals The Euclidean distance between nmax Represents an individual The maximum Euclidean distance between each individual and all other individuals, dist nmin Represents an individual The minimum Euclidean distance to all other individuals.

6. The high-frequency data processing method based on SAS according to claim 5, characterized in that: The position mutation search strategy is used to perform global mutation search on each individual after adaptive information interaction, and the individuals after global mutation search are obtained, including: in, represents the individual after the s-th adaptive information interaction during the t-th training process, Represents the individual after global mutation search Levy represents the random Levy flight factor, α represents the position variation search control factor, Indicates that except for individual Random individuals other than Represents a newly generated random individual, e represents a natural constant, π represents the ratio of pi, r6 represents the sixth random number between (0,1), sin represents the sine function, and T represents the preset maximum number of training times.

7. The high-frequency data processing method based on SAS according to claim 1, characterized in that: The high-frequency data processing model is deployed in the cloud computing center. After collecting the data features of real-time high-frequency data through the statistical analysis system SAS, it is transmitted to the cloud computing center for processing. The data analysis tags corresponding to the data features of the real-time high-frequency data are identified to obtain the high-frequency data processing results, including: Deploy high-frequency data processing models in cloud computing centers; Collect real-time high-frequency data through the statistical analysis system SAS, and extract data features of the real-time high-frequency data through the statistical analysis system SAS; The data features of real-time high-frequency data are transmitted to the cloud computing center, and the cloud computing center schedules the deployed high-frequency data processing model to identify the received data features and obtain the output of the high-frequency data processing model; A high-frequency data processing result is determined according to the output of the high-frequency data processing model.

Images